Sanjay P C created LIVY-713:
-------------------------------
Summary: Yarn Resource Held up after task fails.
Key: LIVY-713
URL: https://issues.apache.org/jira/browse/LIVY-713
Project: Livy
Issue Type: Bug
Components: API
Affects Versions: 0.5.0
Reporter: Sanjay P C
Attachments: airflow.PNG, livy.PNG, yarn-resource.PNG
Overview :
Here we are using airflow as scheduler to run task and Livy is used as
interface to submit spark jobs from airflow to AWS EMR cluster.
Issue :
Here assuming airflow as initiated a job/task which inturn sends message to
livy to submit spark jobs. Now YARN has allocated resources for the submitted
task. If a task is failed due to a bad record / missing file / any exception,
then airflow marks task as Failed but task submitted by livy is stated as idle
and YARN states it has UNKNOWN. Due to this the resources are being held up and
any other task submitted to yarn will in a UNKNOWN state.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
